About me
I am a Ph.D. candidate at Penn State, fortunate to be advised by Prof. Jinyuan Jia, where I learned how to conduct research. I also spent half a year as a Student Researcher at ByteDance, working with Dr. Junjie Liang, Dr. Bochuan Cao and Dr. Xun Chen, where I learned how to turn research into real-world systems.
I received my B.Eng in Computer Science from Wuhan University in 2024.
Research Interest
I build automated red-teaming systems for LLMs and AI agents. My current work focuses on self-improving red-teaming agents that refine their own from experience and human feedback, along with the evaluation environments and sandboxes needed to run and assess agents under realistic settings. I also worked on prompt injection and AI Safety/Security, including Open-Prompt-Injection (USENIX Security 2024), PoisonedRAG (USENIX Security 2025), PIArena (ACL 2026), and PISanitizer (EMNLP 2026).
Experience
- Student Researcher / Research Scientist Intern, ByteDance, May 2026 - October 2026, San Jose, CA.
Publications
* Equal contribution
- Runpeng Geng, Yanting Wang, Chenlong Yin, Minhao Cheng, Ying Chen, Jinyuan Jia. PISanitizer: Preventing Prompt Injection to Long-Context LLMs via Prompt Sanitization (Main Conference), In EMNLP, 2026.
- Runpeng Geng*, Chenlong Yin*, Yanting Wang, Ying Chen and Jinyuan Jia. PIArena: A Platform for Prompt Injection Evaluation (Main Conference), In ACL, 2026.
- Yanting Wang, Runpeng Geng, Ying Chen, Jinyuan Jia. AttnTrace: Contextual Attribution of Prompt Injection and Knowledge Corruption, In IEEE S&P, 2026.
- Chenlong Yin, Runpeng Geng, Yanting Wang, Jinyuan Jia. PISmith: Reinforcement Learning-based Red Teaming for Prompt Injection Defenses, In COLM, 2026.
- Yanting Wang, Runpeng Geng, Jinghui Chen, Minhao Cheng, Jinyuan Jia. TASO: Jailbreak LLMs via Alternative Template and Suffix Optimization, In SeT-LLM Workshop @ KDD, 2026.
- Wei Zou*, Runpeng Geng*, Binghui Wang, and Jinyuan Jia. PoisonedRAG: Knowledge Corruption Attacks to Retrieval-Augmented Generation of Large Language Models, In USENIX Security Symposium, 2025.
- Yanting Wang*, Wei Zou*, Runpeng Geng, Binghui Wang, and Jinyuan Jia. TracLLM: A Generic Framework for Attributing Long Context LLMs, In USENIX Security Symposium, 2025.
- Yupei Liu, Yuqi Jia, Runpeng Geng, Jinyuan Jia, and Neil Zhenqiang Gong. Formalizing and Benchmarking Prompt Injection Attacks and Defenses, In USENIX Security Symposium, 2024.
Preprints
- Runpeng Geng, Yanting Wang, Ying Chen, Jinyuan Jia. UniC-RAG: Universal Knowledge Corruption Attacks to Retrieval-Augmented Generation.
- Yanting Wang, Chenlong Yin, Runpeng Geng, Jinyuan Jia. Agent Against Agent: An Agentic System for Automatic Prompt Injection Red Teaming.
- Yanting Wang, Wei Zou, Runpeng Geng, Jinyuan Jia. AgentWatcher: A Rule-based Prompt Injection Monitor.
Open Source Projects
- PIArena - lead contributor and first author
- PISanitizer - lead contributor and first author
- PoisonedRAG - lead contributor and co-first author
- Open-Prompt-Injection - co-author
